Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Look up keyword
Like this
0Activity
0 of .
Results for:
No results containing your search query
P. 1
User-Based Motion Sensing and Fuzzy Logic for Automated Fall Detection in Older Adults

User-Based Motion Sensing and Fuzzy Logic for Automated Fall Detection in Older Adults

Ratings: (0)|Views: 10|Likes:
Published by Urmi Modi

More info:

Published by: Urmi Modi on Oct 08, 2012
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

10/08/2012

pdf

text

original

 
683
TELEMEDICINE AND
e
-HEALTHVolume 13, Number 6, 2007© Mary Ann Liebert, Inc.DOI: 10.1089/tmj.2007.0007
Original Research
User-Based Motion Sensing and Fuzzy Logic forAutomated Fall Detection in Older Adults
PATRICK BOISSY, Ph.D.,
1,2
STÉPHANE CHOQUETTE, B.Sc.,
1,2
MATHIEU HAMEL, M.Sc.,
1
and NORBERT NOURY, Ph.D.
1,3
ABSTRACTMore than one third of community-dwelling older adults and up to 60% of nursing home res-idents fall each year, with 10–15% of fallers sustaining a serious injury. Reliable automatedfall detection can increase confidence in people with fear of falling, promote active safe liv-ing for older adults, and reduce complications from falls. The performance of a 2-stage falldetection algorithm using impact magnitudes and changes in trunk angles derived from user-based motion sensors was evaluated under laboratory conditions. Ten healthy participantswere instrumented on the front and side of the trunk with 3D accelerometers. Participantssimulated 9 fall conditions and 6 common activities of daily living. Fall conditions were sim-ulated on a protective mattress. The experimental data set comprised 750 events (45 fall eventsand 30 nonfall events per participant) that were classified by the fall detection algorithm aseither a fall or a nonfall using inputs from 3D accelerometers. Significant differences for im-pacts recorded, trunk angle changes (
 p
Ͻ
0.01), and detection performances (
 p
Ͻ
0.05) werefound between fall and nonfall conditions. The proposed algorithm detected fall events dur-ing simulated fall conditions with a success rate of 93% and a false-positive rate of 29% dur-ing nonfall conditions. Despite a slightly superior identification performance for the ac-celerometer located on the front of the trunk, no significant differences were found betweenthe two motion sensor locations. Automated detection of fall events based on user-based mo-tion sensing and fuzzy logic shows promising results. Additional rules and optimization ofthe algorithm will be needed to decrease the false-positive rate.
1
Research Centre on Aging, Sherbrooke Geriatric University Institute, Sherbrooke, Quebec.
2
Department of Kinesiology, FEPS, Université de Sherbrooke, Sherbrooke, Quebec.
3
TIMC-IMAG, Université Joseph Fourier, Grenoble, France.
INTRODUCTION
F
ALLS ARE A MAJOR PUBLIC HEALTH CONCERN
and one of the greatest obstacles to inde-pendent living for older adults. More than onethird of community-dwelling older adults andup to 60% of nursing home residents fall eachyear.
1
The incidence of falls rises steadily withadvancing age and gets even worse amongnursing home residents, where multiple falls
 
(more than 3 per year) are more frequent andinjurious.
2
For older adults who live alone, be-coming incapacitated and unable to get helpimmediately after a fall is also a common ex-perience. In a population-based study, Gurleyand colleagues
3
showed that in people 65 orolder who lived alone, the annual risk of beingfound helpless or dead at home by paramedicsis 3.2%. Weakness or the inability to get up andfalls were the most frequently cited precipitat-ing causes of incapacitation among the sur-vivors found. Nevitt and colleagues
4
reported539 falls after a 1-year prospective cohort studyon 189 ambulatory adults over 60 years of agewho had fallen once in the previous year. Fiftypercent of the fallers reported being unable toget up without assistance in 41% of the falls.Even though most falls have limited physi-cal consequences and are not functionally lim-iting,
4
they can lead to a loss of confidence andthe development of a fear of falling.
5
Commu-nity-based studies of independently livingolder adults have estimated that 25–50% of thispopulation have a fear of falling.
6–9
Fear of falling is a complex problem that can affect not just people who have fallen but also those whohave not. It increases with age, especially forthose who have had a fall in the past, and canalso lead to other negative outcomes, such asfunctional decline,
10
 balance deterioration,
11
and decreased social contact.
12
Fear of falling has generated an industrymarketing automatic alarm-and-notificationsystems, commonly called personal emergencyresponse systems (PERS). PERS take manyforms, but the common thread is that the per-son at risk is equipped with an electronic de-vice carried or worn that, after being activated by the user, can inform a central system whena potentially adverse event occurs. Studies onthe use of PERS as a substitute for in-home su-pervision have found them to be cost-effective inthe context of hospital utilization rates and in-stitutionalization among community-dwellingelders.
13,14
Moreover, they offer a sense of security to family members caring for olderadults living alone and can alleviate their per-ceived burden.
15,16
Although PERS constituteimportant technological adjuncts to home care,they require the direct intervention of the per-son to activate the signal indicating an emer-gency. In the case of a fall-related emergency,this can be impractical when the person is ly-ing unconscious on the floor,
1
is unable to ac-tivate the button, or suffers from cognitive im-pairments. Reliable automated fall detection,coupled with an appropriate help responsefrom a community alarm center, can increaseconfidence in people with fear of falling, pro-mote active safe living for elders, and reducecomplications from falls.
17,18
Over the years,many technological approaches have been de-veloped to address this issue. They can be di-vided into two groups: environmental motionsensing based on “exosensors” positioned inthe individual’s home
19–21
and user-based mo-tion sensing with “endosensors” worn by theindividual. Environmental sensing uses algo-rithms to determine whether, according to in-puts from image sensors (cameras, single-ele-ment passive infrared sensors, pyroelectricinfrared sensor arrays) positioned in the envi-ronment, a fall has occurred in a known vol-ume.Numerous groups have studied fall detec-tion employing user-based motion sensingwith wearable devices.
17,22–27
Typically, mo-tion sensors embedded with a microcontrollerrecord the kinematics of body segments anddetect fall events from the recorded signals us-ing expert systems. Once a fall event is de-tected, an alarm is sent through a radiofre-quency link with a communication deviceconnected to a PERS. Variations on this designhave appeared in various commercial formsand/or are still under development. All of these designs use similar motion sensors(mostly 2D or 3D accelerometers) but detect fallevents differently, using (1) kinematic signalsrecorded at different locations on the body(waist, neck, midtrunk, arm); or (2) different in-puts and processing approaches from therecorded signals (impact, energy of the impact,change in position, etc.); or (3) different expertsystems (Boolean logic, fuzzy logic, neuralnets) to infer from these inputs that a fall hasoccurred. Although numerous designs and sys-tems for automated fall detection based on mo-tion sensing with wearable devices have beenproposed and in some case commercialized,published experimental data on their perfor-mance in real life or under laboratory condi-
BOISSY ET AL.684
 
tions are scarce and generally of poor scientificquality.The objectives of this study were (1) to ex-plore the validity of a two-stage fall detectionalgorithm based on fuzzy logic using experi-mental data collected from simulated falls withhuman participants, and (2) to determine theinfluence on the detection performance of thealgorithm considering inputs from 3D ac-celerometers positioned at 2 locations on thetrunk. The fall detection algorithm and its per-formances during simulated conditions of fallsand non-falls are described below.
MATERIALS AND METHODS
Fall detection algorithm
The proposed fall detection algorithm is based on the identification of 2 separate eventscommon to most falls: impact amplitudes whenthe body decelerates as it makes contact withthe ground, and relative changes in trunk po-sition associated with those impacts (Fig. 1). Wehypothesize that the combination of these twoevents as inputs for a fall detection algorithmwill provide strong identification performanceof simulated falls and eliminate most of thefalse-positive situations (i.e., where a fall didnot occur but the fall algorithm identifies one).The schematic of how these two inputs are usedfor fall detection is illustrated in Figure 2.Briefly, detected impacts are classified using afuzzy logic. Then Boolean logic combines thelevel of impact and the changes in trunk anglesto make a final decision (fall or no fall).
Impact classification using fuzzy logic.
Fuzzylogic is typically used to ease the modeling andprocessing of complex tasks by allowing un-certain states. For example, driving a car would be considered quite complex if the driverwould need to calculate the exact speed of thecar as well as the precise force to apply on the brake and throttle. In this particular scenario,
AUTOMATED FALL DETECTION IN ELDERLY685
FIG. 1.
Typical accelerations pattern during a fall. Thealgorithm detects the highest peak of acceleration (ab-solute value) in (
B
) and considers this event as the impacton the ground. The mean value of acceleration over a 1-second period before (
A
) and after (
C
) an impact is de-tected and used to calculate changes in static trunk angle before and after the fall.
FIG. 2.
Schematic of input and expert system used for fall detection algorithm.

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->